Building smarter networks [Day Two]

SearchNetworking ANZ's Consulting Editor Ian Yates continues his interview with Dimension Data's Paul Wilkinson and focuses on the future of network security.

Ian Yates: Obviously people want the network as fast as they can afford or need, but are they also looking for ways to make it even more secure?

Paul Wilkinson: Yes. From what I see, apart from specific needs such as multimedia or medical imaging or something like that, most people when you ask them do you want it faster or more secure, would probably go for more secure.

Ian Yates: What about making the network more resilient and redundant. Is that something you're being asked for? Maybe 1GB is fast enough, but I don't want it to go down ever.

Paul Wilkinson: Yes, certainly in the core network, that's something we've been doing - it's quite common place and has been for a number of years. We would put in redundant core switches with dual attached ISP services so there's a whole lot more resiliency there. What happens on the edge, with floor switches, sometimes it'll be nothing, it'll be like well if we lose 30 or 60 machines it's not the end of the world and in two hours we'll have the spare parts. But increasingly there are things like redundant power supplies, redundant processor cards - it becomes a bit of a cost trade-off.

What does it cost me to have this floor of people not working? It can happen though with converged communications that once your telephony depends upon that same Ethernet switch, the stakes can be raised because while they might say we can live without email for two hours, we can't live without having the phones for two hours or something like that.

Ian Yates: Yeah, well we can't even ring the tech to ask for a repair!

Paul Wilkinson: That's right. If we don't worry about putting UPSs on the smaller switches because if the floor is blacked out, well no one can use their computers anyway, but if suddenly the phones are hanging off that, and now nobody can make any phone calls. So there has to be UPSs everywhere and then we're getting into management questions. Every two years the batteries die in the UPS, so unless you're monitoring that somewhere, you won't know until the power blacks out that your battery has died.

It's important to know - even if you've got redundant power supplies, but one of them has failed. So we need management so you know one of them has failed rather than relying on someone to notice next time they're walking past. It's probably not going to happen if it's locked up in a floor cupboard somewhere.

Ian Yates: Is anyone doing things like including UPS management in their routing and switching boxes?

Paul Wilkinson: I haven't seen it any switching boxes, but most of those things are all SNMP enabled and have been for a number of years. You can plug all that into your Ethernet and get monitoring and APC have systems which can even monitor how much current you are drawing in a particular rack.

I've seen customers start out with all good intentions and they buy the whole monitored thing and then they go and buy some pieces of monitoring software, whatever brand it is, and it sits on the shelf. Or someone gets around to setting it up and then no one maintains it so after a while lots of the things that were monitored aren't there any more so there's all sort of red alarms, so someone just turns it off.

That's why we offer managed services around that sort of idea. Yep, someone's usually monitoring it. You probably won't know how to do it, so you're going to set it up once and then forget it, so you can outsource that.

Ian Yates: In a sense it's a bit like an insurance policy. You don't really want to pay your own staff to sit and stare at a screen all day when it never goes wrong, just in case it goes wrong.

Paul Wilkinson: Yes, and running out of disk space on a server, someone's going to notice because someone's going to say I can't save my file. You want to be a bit more proactive than that hopefully, but monitoring the servers is really pretty easy because you can get a gut feel for it, I mean something's running a bit slow, loss of memory. It's all of those forgotten peripherals that can be monitored which aren't. It's pretty straight forward to receive an SNMP trap for something that says my power supply died, and then act upon it. But if you haven't configured it, you'll never know.

Ian Yates: I suppose it's a bit like if you own the latest model car, if you don't take it to your dealer every six months, nobody will be reading all the little messages it's putting out saying I'm about to drop dead.

Paul Wilkinson: That's right. You'll get in one morning and that little orange engine light will come on and the whole thing won't start. But you could have dealt with something simple six months ago and avoided the whole problem.

< Back to Day One


Dig deeper on Network design